Machine learning to solve the Travelling Salesman Problem
My Undergraduate Research project
I created an interface in javascript and php that allows users to solve the travelling salesman problem manually as a game, and using a genetic algorithm that optimizes a customized neural network.
The code was written to run both linearly and in parallel, by using web workers, to improve runtime.
I was able to find better results than most of my peers and have a fun UI to go with it. UI uses a html canvas and javascript to visualize solving the problem in real time. URL query parameters can be used to select pre-recorded pointsets and neural network weights, or they can be set to zero or ommited to be randomized on page load.
Human Solver (game) https://167.99.146.165/travellingsalesman.php?x=14&y=1
Linear Solver https://167.99.146.165/travellingsalesman.php?x=1&y=1
Parallel Solver https://167.99.146.165/parallelsalesmancleaned.php